COMMENT ⊗   VALID 00003 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	Mr. Howard Boyer
C00007 00003	morave.rev[f86,jmc]	Review of Moravec's Mind Children
C00010 ENDMK
C⊗;
Mr. Howard Boyer
Harvard University Press
79 Garden Street
Cambridge, MA 02138

Dear Mr. Boyer:

	I think you should publish Hans Moravec's {\it Mind Children}.
The main reson is that it has a large number of ideas that many
readers will find interesting.  Moravec's presentation is quite
clear.  I think that most readers, like me, will find some of
the ideas convincing and others not.  There isn't a single topic
to cover the essence of.  Futurism is much to open-ended for that,
but this book will stretch many people's imaginations.  I would expect the
book to be popular among people with a scientific or technical bent
 interested in futuristic ideas.  There are many at all levels of
scientific knowledge.

	Here are some things that should be fixed.

	1. The later chapters are more interesting than the earlier
chapters.  I found the computer history rather dull, because it
was a rehash of previously presented material.  Perhaps a beginner
would find it more interesting.  The projections of future increases
in computer speed are also rather well known material.

	2. Although the references are incomplete, I suspect that
he doesn't intend to provide references as complete as he should.
It is important that readers be able to track down ideas of other
people that Moravec refers to.

	3. There is too much duplication between the autobiographical
material in the introductory chapter and the same information in
the robotics chapter.

	4. While I like the skyhook ideas, there isn't very much
connection between them and the other material.

	5. He should go over the book with a spelling checker, but
it won't tell him that von Neumann's name has two n's at the end.
On page 128, it should be Carl Friedrich Gauss.  Fermi, p. 178, shouldn't
be characterized as a ``Manhattan project scientist''.  He was
primarily a basic researcher, and his widow would be offended.
I think that while ENIAC may have been designed at the University
of Pennsylvania, it was built at the Aberdeen Proving Grounds in
Maryland.

	6. The material on complexity on p. 198 has several errors,
which should be fixed or the topic omitted.  For example, matrix
multiplication doesn't have complexity $O(n↑3)$.  As I recall, its
complexity is at least $O(n↑2)$, but it is known to have complexity
less than $O(n↑{2.8})$.  This is easily shown from the fact that
2 by 2 matrices can be multiplied with 7 multiplications.  Not
all NP problems are equivalent --- only NP-complete problems.
None of these mistakes are essential to Moravec's main points,
but they will irritate people.  He can easily look it up.

	7. On p. 149, he should credit the science fiction story
Waldo with the first idea of a force amplifying hand.  If I remember
correctly, the story is by Waldo somebody.

	8. Many of the ideas for realizing nanocomputers, etc. in
the first part of the book involve wishful thinking.  This isn't
sufficiently acknowledged.

	9. Perhaps he should elaborate the ``What am I?'' chapter.

Sincerely,

P.S. I'll take the cash.
morave.rev[f86,jmc]	Review of Moravec's Mind Children

needs full references, so that readers can follow his tracks

there is duplication between the introductory autobiographical
chapter and some of the others.

there should be a reference to at least one standard AI text

too much computer history

I like the skyhook stuff, but its connection with the rest of 
the book isn't great

I find the extrapolations of computer power into the future
rather dull, though not inaccurate.

it should be more frank about its wishful thinking

128 Carl Friedrich Gauss

he might be encouraged to elaborate the What am I chapter

134 Wasn't ENIAC built at Aberdeen Proving Grounds, Maryland?
	von Neumann has two n's

149 credit the science fiction Waldo

178 Fermi wouldn't like to be characterized as "Manhattan Project physicist".
He was primarily famous as for basic research.

198 Multiplying two matrices does not have complexity n↑3.  n↑2 is
a known lower bound, and n↑2.8 is quite simply achieved by building
up from a procedure that multiplies 2 by 2 matrices with 7 multiplications.

The traveling salesman problem is O(k↑n), I think.

It isn't true that all NP problems are interconvertible.  It's all
NP-complete problems that are interconvertible.

214 hierarchy, inevitability